Erweiterte Suche

Variational Methods for Joint Motion Estimation and Image Reconstruction

Thèse - Institute for Computational and Applied Mathematics University of Muenster - june 2015
Télécharger la publication : digitalFinal.pdf [14.6Mo]  
In this thesis we present different techniques for motion estimation from image sequences and combine them with image reconstruction. The main body of this work is divided in two modeling and one application part. The first model is a variational approach for motion estimation from image sequences. Mathematical background as well as different models for motion estimation are presented. We illustrate the numerical realization based on a primal-dual framework and evaluate our model towards different types of motion.\ In the second main part we connect the field of motion estimation to the task of image reconstruction. Here, we deduce variational models for joint motion estimation and image reconstruction, prove existence of minimizers and present primal-dual schemes for the numerical implementation. The later numerical evaluation illustrates benefits of the joint framework and contains promising results for temporal inpainting. The application part divides into chapters about image segmentation, where Geodesic Active Contour and Chan-Vese models are applied to images of vertebra and intracellular flows. In the context of intracellular flow, we later apply our joint models to denoise image sequences and estimate their underlying motion simultaneously.

Références BibTex

@PhdThesis{Dir15,
  author       = {Dirks, H.},
  title        = {Variational Methods for Joint Motion Estimation and Image Reconstruction},
  school       = {Institute for Computational and Applied Mathematics University of Muenster},
  month        = {june},
  year         = {2015},
  keywords     = {Image Processing, Image Reconstruction, Motion Estimation, Joint Image Reconstruction and Motion Estimation, Total Variation, Optical Flow, Microscopy Imaging, Biomedical Imaging, Temporal Inpainting, Variational Methods, Primal-Dual Methods},
  url          = \{/2015/Dir15},
}

Autres publications dans la base


Impressum | © 2007 FB10 WWU Münster
Universität Münster
Schlossplatz 2 - 48149 Münster
Tel.: +49 (251) 83-0 - Fax: +49 (251) 83-3 20 90
E-Mail: